ChatGPT-4.0 or DeepSeek-V3? Comparative analysis of answers to the most frequently asked questions by total knee replacement candidate patients.

BACKGROUND

Total knee arthroplasty (TKA) is a surgical intervention that significantly improves patients' quality of life, but the preoperative process can cause uncertainty, anxiety, and a lack of information. In recent years, artificial intelligence (AI)-powered chatbots and large language models have begun to play important roles in patient information processes in the healthcare field. In this study, the answers given by chat generative pretrained transformer (ChatGPT)-4.0 and DeepSeek-V3 AI programs to the 10 most frequent questions about TKA asked by patients before surgery were compared, and the effectiveness of AI in the patient information process was analyzed with the evaluations of orthopedists.

METHODS

Using Google Trends, patient forums, and clinical experiences, the 10 questions that TKA patients are most curious about in the preoperative, peroperative, and postoperative periods were determined. These questions were directed to ChatGPT-4.0 and DeepSeek-V3, and the answers were recorded. Five orthopedists (minimum 5 year surgical experienced) evaluated the answers using a Likert scale (1-5) according to criteria such as scientific accuracy, explanatory power, understandability for the patient, and detailed content.

RESULTS

The mean Likert score of ChatGPT-4.0 (4.7 ± 0.2) was found higher than the mean Likert score of DeepSeek-V3 (3.5 ± 0.3) (P < .05). ChatGPT-4.0 provided more comprehensive and detailed information, while DeepSeek-V3 provided superficial answers, especially in the answers to questions such as "life of the prosthesis," "postoperative complications," and "return to daily activities."

CONCLUSION

Our study showed that ChatGPT-4.0 is more effective than DeepSeek-V3 in terms of patient information regarding total knee replacement. It is emphasized that AI-supported systems are a fast and accessible source of information for patient education; however this information must be inspected by medical authorities for accuracy. Future studies should be conducted with larger patient populations, to increase the reliability of AI-based patient information systems and ensure their integration into clinical practice.

LEVEL OF EVIDENCE

Level 5.